7225056 Multiple Reuse Patterns For Frequency Planning In Gsm Networks

MULTIPLE USE PATT FOR FREQUENCY PLANNING IN GS Stefan Engstrom, Thomas Johansson, Fredric Kronestedt, Magnus Larsson, Stefan Lidbrink, H&an Olofsson Ericsson Radio Systems AB S-164 80 Stockholm Sweden

-

Abstract Providing high capacity in GSM networks at low costs using existing macrocells is of increasing importance in the near future due to the competition between operators. This paper shows that by applying frequency hopping in combination with an advanced frequency planning method, Multiple Reuse Patterns (MRP), very high traffic levels in the existing macrocells can be supported. Field experience from live networks shows that an average frequencyreuse factor as low as 7.5 is possible without negatively affecting the network quality. Thus, the network capacity can be doubled compared to a non hopping network with reuse 4/12 using 10 MHz frequency spectrum. In a macrocell, it is possible to carry as much as 43 Erlang at 2% blocking.

I. INTRODUCTION The cellular market has experienced an enormous subscriber growth in the recent years. Today, GSM networks in more than 100 countries serve approximately 65 million subscribers. It is of significant importance for the network operators to support high capacity in their networks at minimum costs due to the increasing competition [ 11. There are several ways to increase capacity from a cell planning point of view. Methods in use today include cell split, overlaidunderlaid cells and hierarchical cell structures. In general, these methods can be divided into two groups, where one requires addition of new cell sites, while the other only implies installation of new transceivers in already existing base stations. Deployment of new cell sites is often a fairly slow process, due to site acquisition problems. This, together with the cost of new site equipment makes this option less efficient from a cost and implementation perspective. The alternative method, to reuse existing cells and only adding transceivers, is thus an attractive option. Addition of transceivers to existing cells can be facilitated by applying radio network features such as e.g. overlaid underlaid cells, frequency hopping, power control and Discontinuous Transmission (DTX). These features reduce andor change the characteristics of the network interference so that a tighter frequency plan can be applied and hence more transceivers can be added. In this paper, a solution for providing high capacity in GSM using existing macrocells is highlighted. The solution, known as Multiple Reuse Patterns (MRP) [ 1-31, uses tight frequency reuse in combination with frequency hopping. The paper also provides results from MRP field trials. Finally, discussions

0-7803-4320-4/98/$5.00 0 1998 E E E

regarding 1/3 frequency reuse [4-51, a similar solution, are included. 11. FREQUENCY HOPPING Increasing network capacity by tightening the frequency reuse results in a heavily interfered radio environment. This makes it more difficult to produce a frequency plan of good quality. In the end, interference managing techniques such as frequency hopping, power control and DTX are required in order to secure the quality in the network. This paper considers only the use of frequency hopping. With frequency hopping, frequency diversity will occur, which balances the quality between slow and fast moving users. This implies that a cell planning margin for fast fading (Rayleigh fading) is not needed. Thus, an approximate coverage gain equal to the fast fading margin can be achieved from the frequency diversity effect. Today, cell planners typically use 3 dB as the fast fading margin. Furthermore, frequency hopping also introduces interference diversity [4]. Two aspects of interference diversity combine to improve performance. Strong interferers are shared between different users, which is often referred to as intetfierence averaging. In addition, the time varying interference as such increases the interleaving efficiency and thus improves receiver performance. Altogether, a frequency plan with tighter reuse can be implemented in a frequency hopping network, resulting in improved capacity compared to a non hopping network. 111. MAXIMIZING INTERFERENCEDIVERSITY Interference diversity due to frequency hopping can be seen as a reduced correlation of the interference signals experienced for consecutive bursts. Figure 1 illustrates this signal correlation decrease for three simplified scenarios, in which the uplink of a connection in cell A is interfered by mobile stations in co-channel cells. Cell A is assigned frequencies 1 and 10 in all scenarios. In the leftmost scenario without frequency hopping, the connection in cell A continuously uses frequency 1, and therefore the interference I arises from the same user in cell B all the time. The correlation of the interference signal on consecutive bursts is thus high. If the connection has bad quality, an improvement can only be made if the co-channel cell stops transmitting on this frequency or if the connection in 2004

VTC ‘98

No frequency hopping

Frequency hopping with traditional planning

I , IO

Frequency bopping without fixed frequency groups I. 6

@8

A. Band split

I , IO

1, IO

MS on frequency 1 in cell A.

MS on frequency 1 and 10 in cell A.

of tightness and use frequency hopping to combine them into an average reuse. The goal is to deploy as many transceivers as possible in existing cells to minimize the number of costly new sites. In this paper, only MRP using baseband frequency hopping is considered.

MS on frequency 1 and IO in cell A

Figure 1. Example of the interference diversity effects of different frequency hopping strategies.

The first step with MRP is to split the available frequency spectrum into different bands. One band is the BCCH band which means that a frequency used as a BCCH frequency in one cell will not be used as a TCH frequency in another cell and vice versa. The reasons for reserving unique BCCH frequencies are: Trafic independent BSZC decoding pelformance: When the mobiles are trying to decode the BSIC (Base Station Identity Code) on the SCH (Synchronisation Channel), the performance will not be affected of the traffic load in the network. The reason is that the traffic assigned to the TCH frequencies will never disturb any BCCH frequency on which the SCH is mapped. BSIC decoding is very important for the handover performance. Poor handover performance could lead to increased number of dropped calls.

cell A is handovered (by an inter-cell or intra-cell handover). The middle scenario shows the traditional frequency hopping case, in which regular frequency groups are used. The connection in cell A hops on two frequencies (1 and lo), which are both used in cell B as well. Consequently, the interference origin will alter between two users in cell B, causing the two interference signals lI and Z2. Since the strength of ll and Z2 may be rather different, the interference signal correlation may be fairly low for consecutive bursts. In other words, the interference diversity has increased compared to the nonhopping case. Finally, in the rightmost scenario, an irregular frequency plan is applied together with frequency hopping. Typical for this case is that there are no fixed sets of frequencies used in a cell and its co-channel cells. Thus, cell B is only a partial cochannel cell of cell A, since they have only one frequency in common. On the other hand, this arrangement creates an increased number of (partial) co-channel cells, in this example represented by cell C. In this case, different bursts of a connection in cell A will be interfered by users in different cells. Thus, consecutive bursts will experience the interference signals Zl and Z2, which generally are totally uncorrelated. Hence, this scenario is superior to traditional planning with regular frequency groups in terms of maximizing interference diversity. The example above indicates that to obtain maximal interference diversity, a frequency plan without frequency groups is preferable. However, such frequency plan has apparent drawbacks, including the extensive re-planning necessary in a continuously evolving network. By applying the MRP technique, it is possible to provide maximal interference diversity and still maintain a structured frequency plan, as will be described in the next section.

Simplified neighbor cell list planning: The number of possible neighbor cell frequencies decreases with a separate BCCH band. A simple strategy where all frequencies except the own BCCH frequency are included in the neighbor cell list can be used. Using all available frequencies as BCCH frequencies may result in longer neighbor cell lists, which has negative impact on handover performance [6]. Full gain from power control and DTX:Only TCH frequencies can use DTX and power control in the downlink. With a dedicated BCCH frequency band, full gain from power control and DTX is achieved in the downlink [7]. This is not the case if the BCCH frequencies interfere with the TCH frequencies. Thus, a more aggressive power control approach can be applied. Replanning of TCH frequencies will not affect the BCCH frequency plan: If additional transceivers are to be added to already existing cells, the BCCH frequency plan is not affected (assuming that the combiner spacing requirements are neglected). The only restriction to consider is the adjacent frequency interference. Thus, it is possible to keep the same BCCH plan even if additional transceivers are added to the network. The network operator then knows that if the BCCH frequency plan is good it will remain good, independent of the TCH frequencies.

Iv. MULTIPLEREUSE PATERNS Multiple Reuse Patterns (MRP) is a generic method for achieving high capacity using tight frequency reuse in combination with frequency hopping [l-31. The MRP technique exploits the advantages from frequency hopping in order to increase the capacity. The fundamental idea with MRP is to apply different separate reuse patterns with different degrees

0-78034320-4/98/$5.00 0 1998 IEEE

As a next step, the MRP method implies that the remaining (TCH) frequencies are divided into different bands. Thus, one BCCH and several TCH bands will exist. The main idea with

2005

VTC '98

several TCH bands is to apply different reuse patterns on different TCH transceivers. The first TCH transceiver in all cells will use frequencies from the first TCH band and so on. The reasons for splitting the TCH frequencies into different bands are: Dimension the average frequency reuse according to the transceiver distribution of the network: The transceiver distribution determines the average frequency reuse that can be applied in the network. The average frequency reuse is adjusted according to the maximum number of transceivers needed per cell and the number of cells requiring this number. In this way, the quality can be controlled in an effective manner in the frequency planning process. Small impact on existing frequency plan when adding more transceivers: The TCH band split will limit the required amount of frequency planning work when more transceivers are added. Only the cells with the same number of transceivers or more will be affected if more transceivers are added. For example, adding a fourth transceiver to a cell with three transceivers will only have an influence on the cells with four and more transceivers.

A structured way of frequency planning: It is possible, for instance, to make a frequency plan for the first TCH transceiver without modifying the BCCH plan or the plans for the other TCH transceivers. This structure will aid the frequency planner in hisher work by making it easier to produce a new frequency plan and to detect a bad frequency plan.

B. Frequency Assignmenl The MRP frequency assignment can be exemplified by means of Figure 2, which shows a schematic picture of how the different frequencies can be allocated to an MFW configuration with maximum four transceivers per cell. The example is referred to as a 12/10/8/6 plan. This means that there are 12 BCCH frequencies (frequencies 1,3,...,23), 10 TCH frequencies in group 1 (frequencies 2,4,...,20), 8 TCH frequencies in group 2 (frequencies 22,24,...,36) and 6 TCH frequencies in group 3 (frequencies 25,27, ...,35). Figure 2 shows only the frequency allocations for two cells A and B which have two and four transceivers respectively. Cell A is assigned the BCCH frequency 1 and the TCH frequency 6. Thus, cell A uses baseband frequency hopping over two frequencies. Further, cell B is allocated the BCCH frequency 23 and the TCH frequencies 6, 26 and 35. Consequently, cell B uses baseband frequency hopping over four frequencies. Note that the frequencies defined as BCCH frequencies do not have to be defined as shown in Figure 2. Hence, any frequency from the available spectrum can selected as a BCCH as long as the BCCHRCH separation is fulfilled. There is no need to strictly adhere to the MRF’ technique all the time. If a cell with quality problems exists, it is acceptable

0-7803-4320-4/98/$5.00 0 1998 IEEE

Slow baseband hopping over 2 frequencies ( l e L6)

8 ’e3

Slow baseband hopping over 4 frequencies (1 e 2 3 , 6 2 6 . 3 5 )

v Cell A

v Cell B

Figure 2. An example of frequency planning with MRP.

to solve this problem by changing a frequency in that cell with an “illegal” frequency which initially is used in another transceiver group. However, it is recommended that the MRP structure is followed as much as possible.

C. Tailoring the frequency plan The MRP scheme has been developed to handle the typical case when networks have uneven transceiver distributions. This is important since every cellular network differs in characteristics regarding e.g. cell sizes, number of available frequencies and topography. This means that some cells have many transceivers while other cells have only a few. To understand how the different cells with different number of transceivers experience different frequency reuse situations, an example is shown in Table I. A 12/8/6/4 MRP configuration which totally requires 30 frequency carriers is selected. There are 12 BCCH frequencies and 3 TCH groups each containing 8, 6 and 4 frequencies. In the example, it is further assumed that different fractions of the cells (20%, 30% and 50%)have 2, 3 and 4 transceivers respectively. The average frequency reuse factor experienced by a cell is defined as the total number of frequencies in the groups assigned to the cell divided by the number of transceivers in the cell. The different cells will thus encounter different average frequency reuse factors: 10 for the two transceiver cells, 8.7 for the three transceiver cells and 7.5 for the four TABLE I An MRP example with 6 MHz of spectrum, where cells have an uneven distribution of transceivers. Numberof transceivers/cell Fraction of cells MRPgroups Average frequency reuse

I( 2

I1

20%

(1

1

12/8 (12+8)/2

=10

Diversity gam

2006

)I

I

1

I

30%

I

12/8/6

I

(12+8+6)/3 =87

10

I

small

I

I

I

3

0 50% 12/8/6/4

(12+8+6+4)/4 =7.5

I

I

,I

Actual reuse (upper bound)

I

92

larger

I

I

4 8.9

largest

VTC ‘98

transceiver cells. The “actual” average frequency reuse experienced by a cell may however be sparser, since not all cells are fully equipped. For example, since the third transceiver is only used in 80% of the cells, the actual reuse on that transceiver may be as sparse as 6/0.8=7.5, depending of the geographical distribution of the cells with the third transceiver. The upper bound for the actual average frequency reuse for cells with three transceivers is therefore (12+8+7.5)/3=9.2 (Table I). The gain from frequency hopping increases with the number of frequencies included in the hopping sequence [4]. Cells with a lot of transceivers may experience a tight reuse (which lead to an increased interference level) but this will be balanced with a larger interference diversity gain. The above example illustrates how MRP can be adjusted to the transceiver distribution in the network. That is, the frequency plan is adjusted to the network. It should be further noticed that MRP do not need to be implemented over the entire network. MRP can just be applied in the area where high capacity is needed. It is also possible to use different MRP configurations in different locations of the network.

v. RESULTS FROM FIELD MEASUREMENTS The MRP method is currently being used in more than a dozen networks. This section presents MRP experience from two typical live networks. Figure 3 shows results from a network where MRP were tested and later implemented. The traffic in the area caused high blocking, around 10-15%, before the trial started. The test area was a dense urban area and included 40 cells, all equipped with four transceiverskell. The site-to-site distance was around 500 meters. The major result is that the dropped call performance (TCH drop) was unchanged when the average frequency reuse was decreased from 15 to 8.2. A low TCH drop value reflects a good network performance and a rate of 2% is normally acceptable. The handover performance was the same for all the different MRP plans. In addition, call setup, location updating and paging performance were not changed either over the course of the test. An average reuse of 8.2 was thus possible. The capacity increase that the combination of frequency hopping and MRP provide is different from network to network. The operator in this case had access to 60 frequencies. With a average reuse of 15, four transceivers per cell was possible. This means that around 21 Erlang per cell was feasible according to the Erlang B table with 2% blocking. With an 8.2 reuse, it was possible to have 7 transceivers per cell. This figure results in 43 Erlang per cell, i.e. the capacity was doubled compared to the 15 reuse configuration. Figure 4 includes result from MRP tests in another dense urban area. In this case, the trial area consisted of 36 cells, all with four transceivers. This area had a site-to-site distance of 400-500 meters. Again, the results showed that the call drop

.

0-7803-4320-4/98/$5.00 0 1998 IEEE

TCH Drop [%I

40cells, 4 transceiverslcell Reference = I5 I 15 I 15 I 15 MRPl=13/12/10/10 MRP 2 = 13 I 8 / 8 / 8 MRP 3 = 13 I 8 I 8 I 6 MRP 4 = 13 I 8 I 6 I 6

3.0

I

I

I

I

I

I

I

I Reuse 15 lReuse -11.21 Reuse-9.2

11

I

I

NoMReference

‘

I

’

MRPl

Reuse -8.7

I

MRP3

’

I

’

MRP2

Reuse-8.2

I

MRP 4

Figure 3. Results from a MRP trial where the average reuse was decreased from 15 to 8.2. The test area dense urban environment with 40 cells, all equipped with four transceivers.

performance was more or less the same when going from a frequency reuse of 12 down to 7.5. A different call drop measure is used in this plot, Erlang*Min/Drop, which reflects how many accumulated minutes a call can be maintained before it is dropped. Furthermore, all other performance indicators, such as e.g. handover, paging and call setup performance, showed no degradation when the frequency plan was tightened. In this case, the potential capacity gain was 100% if additional transceivers would be installed as the frequency reuse was tightened. In the implementation phase, instead of using 7 transceivedcell which is possible with a 7.5 frequency reuse, 6 transceiverdcell were used. The saved frequencies were used for implementing microcells in the area. Thus, unique BCCH frequencies could be used in the microcells, as recommended for MRP networks including microcells [ 11. 7.5 reuse is not a lower limit for the reuse. Additional trials have shown that by applying power control and DTX in the downlink, the average reuse can be decreased below 7. Note that the two previously shown examples were not using downlink power control or DTX. Additional capacity can most likely be extracted from the macrocells in these trial areas by tightening the reuse even further. 36 cells, 4 transceiverslcell Reference = 12 I 12 I 12 I 12 MRP 1 = 1 2 1 9 1 9 1 9 MRP 2 = 12 1 8 1 8 I 6 MRP3= 1 2 / 8 / 6 / 6 MRP4= 1 2 1 8 1 6 1 4 I

Erlang*Min/Drop

60

40

1I

2 o t R e u s e 12

Reference

I

I

I Reuse-9.7

MRP 1

Reuse8.5

Reuse8

Reuse7.5

MRPZ

MRP 3

MRP 4

Figure 4. Results from a second MRP trial where the average reuse was changed from 15 to 7.5.In this case, test area located in dense urban environment consisted of 36 cells, each with four transceivers.

2007

VTC ‘98

The experience from live networks has revealed that the frequency reuse limit mainly depends on three factors: the number of available frequencies (with a large amount of frequency spectrum, a tighter reuse can be applied since more frequencies can be included in the hopping sequence which increases the diversity gain); the coverage performance (it is easier to perform the frequency planning when the coverage is good); and the site-to-site distance (it is more difficult to create a good frequency plan when the cells are located close to each other since each cell has many potential neighbours). In the example in Figure 3, MRP was used for boosting the macrocell capacity only. Instead of using all frequencies in the macrocells, some frequencies could be reserved for other cells, e.g. microcells or indoor cells, as was the case in the second example (Figure 4). The traffic capacity of the existing macrocells is in this scenario not doubled, but the new cells (microcellshndoor cells) improve the capacity even more. Using reserved frequencies will also result in easier implementation of the new cells, since no co-channel interference will exist between the macrocells and the new microcellshndoor cells. VI. 1/3 REUSE WITH FRACTIONAL LOADING

be loaded with traffic up to 100%. Full traffic load would result in insufficient user quality. Hence, fractional loading is required, i.e. the traffic load must be limited in order to maintain the quality. The number of frequencies that are used in a cell is larger than the number of simultaneously available traffic resources. Fractional loading can be achieved either by using hardware or software control [8-91. MRP with slightly sparser reuse than 1/3 allow the cells to be fully loaded, and therefore sparser reuse factors are common in current MRP networks. VII. CONCLUSIONS Implementing tight frequency reuse by using Multiple Reuse Patterns (MRP) with frequency hopping in GSM has been proven to be an efficient way to increase the radio network capacity with minimal costs for a network operator. Field tests from live networks show that it is possible to implement an average frequency reuse of 7.5 without jeopardizing the network quality. Features like power control and DTX were not used in the trials. For comparison reasons, a non-hopping GSM network can at its best cope with approximately a 12 reuse in average. With MRP, it is possible to adjust the tightness of the frequency plan according to the transceiver distribution. At the same time, MRP provides a robust frequency plan which is very insensitive to changes, e.g. addition of transceivers.

REFERENCES

Another way of increasing the network capacity by applying tight frequency reuse and frequency hopping is to use 113 reuse on the TCH frequencies 14-51 (the BCCH frequencies must still use at least a 12 reuse). However, 1/3 reuse is just a special case of MRP where TCH frequencies all have equal reuse; an equivalent MRP configuration is e.g. 12/3/3/3/3. There are some advantages and drawbacks with 1/3 reuse compared to the MRP configurations described earlier in this paper. One advantage with 1/3 is an extensive diversity due to the large number of frequencies in the hopping sequence. Furthermore, reduced frequency planning work may be required for 1/3 reuse compared to MRP since only one TCH frequency plan is required, assuming that no new cells are added. More transceivers can also be added in existing cells without additional frequency planning. However, it might also be very difficult to produce a 1/3 frequency plan in an irregular network with off-grid cells. The quality for a problem cell can often not be improved by changing the frequency plan. In most cases, the problem is due to that the cell experiences interference from a large number of other cells. Hence, finding a dominant carrier signal in this area may be a problem. Instead, a new site may have to be added, since it is impossible to solve the problem by changing one or several frequencies. In the MRP case, it is possible to change the frequency plan in a problem cell, since a sparser frequency reuse is applied. A further disadvantage with 1/3 reuse is that the cells cannot

0-7803-4320-4/98/%5.000 1998 IEEE

M. Madfors et al., “High Capacity with Limited Spectrum in Cellular Systems“, in IEEE Communications Magazine, Aug., 1997.

J. Dahlin, “Ericsson’s Multiple Reuse Pattem For DCS 1800”, in Mobile Communications International, Nov., 1996. A. Kolonits, “Evaluating the Potential of Multiple Re-Use Patterns for Optimizing Existing Network Capacity”, IIR Maximizing Capacity Workshop,London, June, 1997. H. Olofsson et al., “Interference Diversity as Means for Increased Capacity in G S M , in Proceedings of 1st EPMCC, Italy, 1995, pp. 97102. J. Wigard et al., “Capacity of a GSM Network with Fractional Loading and Random Frequency Hopping”, in Proceedings of the 7th IEEE PIMRC, 1996, pp. 723-727.

H. Olofsson, S. Magnusson, M. Almgren, “A Concept for Dynamic Neighbor Cell List Planning in a Cellular System”, in Proceedings of the 7th IEEE PIMRC, 1996, pp. 138-142. F. Kronestedt and M. Frodigh, “Frequency Planning Strategies for Frequency Hopping GSM’, in Proceedings of the 47th IEEE VTC, 1997, pp. 1862-1866.

M. Naghshineh and M Schwartz, “Distributed Call Admission Control In MobilelWireless Networks”, in Proceedings of the 6th IEEE PIMRC, 1995, pp 289-293 P. Beming and M. Frodigh, “Admission Control in Frequency Hopping GSM Systems”, in Proceedings of 47th IEEE VTC, 1997, pp. 12821286.

2008

VTC ‘98